The current document is directed to methods and systems that refine anatomical models to sub-voxel resolution. In certain implementations, sophisticated, composite, digital anatomical atlases provide detailed three-dimensional models of the contents of three-dimensional medical images. However, three-dimensional medical images have limited resolutions characterized by a smallest volume, referred to as a voxel, to which an intensity is assigned by the imaging process. The currently disclosed methods employ computed percentages of different types of tissue within voxel volumes to adjust a three-dimensional model of the contents of the voxel volumes to more accurately model the contents of the voxel volumes.
Legal claims defining the scope of protection, as filed with the USPTO.
1. A method that adjusts and refines, to a sub-voxel resolution, anatomical information at a sub-voxel dimensional scale in order to incorporate intensity information, contained in a primary medical image that contains multiple voxels and n additional medical images that each contains multiple voxels, into the anatomical information, the method comprising: receiving references to the primary medical image and the n additional medical images; for each voxel of a set of voxels within the primary medical image that each corresponds to a voxel volume within the anatomical information, determining a number and types of different tissues in the voxel volume, when there is sufficient intensity data associated with the voxel, in one or more of the primary medical image and the n additional medical images, to estimate a fractional tissue content of the voxel, estimating the fractional tissue contents of the voxel from the intensity data, determining a fractional tissue contents of the voxel volume from the anatomical information, and when the determined fractional tissue contents of the voxel differs from the fractional tissue contents of the voxel volume by at least a threshold difference, adjusting the anatomical information at the sub-voxel dimensional scale within the voxel volume so that the fractional tissue contents of the voxel differ from the fractional tissue contents of the voxel volume by less than the threshold difference; and adjusting the anatomical information at the sub-voxel dimensional scale within any additional voxel volumes containing misaligned portions of features that span multiple voxel volumes so that the features that span multiple voxel volumes are aligned at the boundaries between adjacent voxel volumes.
2. The method of claim 1 wherein the anatomical information is contained in, or generated from, one or more digital anatomical models which may include one or more composite digital atlases that each includes data for many different imaging subjects.
3. The method of claim 2 wherein the primary medical image is input to one or more digital atlases or to an anatomical-information generator that generates anatomical information from one or more digital atlases.
4. The method of claim 3 wherein the primary medical image is input to one or more digital atlases; wherein, in response, the one or more composite digital atlases; search through the data for the different subjects to identify a number of model volumes that most closely represent features within the primary medical image, extract features from these model volumes, and uses the extracted features to generate a result model volume that closely models and corresponds to the primary medical image; and wherein, multiple result model volumes are generated, generating a final result model volume from the multiple result model volumes.
5. The method of claim 3 wherein the primary medical image is input to the anatomical-information generator; wherein, in response, the anatomical-information generator generates features from anatomical information contained in the one or more digital atlases, and uses the generated features to, in turn, generate a result model volume that closely models and corresponds to the primary medical image.
6. The method of claim 3 wherein the one or more composite digital atlases employ positioning, orientation, scaling, and warping operations to generate the result model volume.
7. The method of claim 1 wherein the fractional tissue contents of the voxel volume is directly computed by determining the portion of the voxel volume corresponding to each tissue type.
8. The method of claim 1 wherein, when the voxel volume contains tissues of only two types, the fractional tissue contents of the voxel is computed from an intensity associated with the voxel in the primary medical image and the fractional tissue contents of the voxel volume.
9. The method of claim 8 wherein the fractional tissue contents of the voxel is computed by solving a system of two linear equations, including an equation in which the sum of two terms, each term the product of a mean intensity for a tissue type and the fractional tissue content of the voxel for that tissue type, is equal to the intensity associated with the voxel and including an equation in which the sum of the fractional tissue content for each tissue type in the voxel is equal to 1.0.
10. The method of claim 8 wherein a fractional-tissue-content range for each tissue in the voxel is computed by determining, from tissue intensity profiles, an intensity range for each tissue type; using the minimum value in the intensity range for a first tissue type and the minimum value in the intensity range for a second tissue type, determining a maximum fractional tissue content for the first tissue type; using the maximum value in the intensity range for the first tissue type and the maximum value in the intensity range for the second tissue type, determining a minimum fractional tissue content for the first tissue type; and determining corresponding minimal and maximal fractional tissue contents for the second tissue type.
11. The method of claim 1 wherein, when the voxel and voxel volume contain m types of tissue, where m is less than or equal to n+2, the fractional tissue contents of the voxel is computed from an intensity associated with the voxel in the primary medical image and in m−2 of the n additional medical images by solving a system of m linear equations.
12. The method of claim 1 wherein, when the voxel and voxel volume contain m types of tissue, where m is less than or equal to n+2, a fractional-tissue-content range for each tissue type in the voxel is computed by for each tissue type, using the minimum value in the intensity range for the tissue type and the minimum values in the intensity ranges for the remaining m−1 tissue types, determining a maximum fractional tissue content for the tissue type; and using the maximum value in the intensity range for the tissue type and the maximum value in the intensity ranges for the remaining m−1 tissue types, determining a minimum fractional tissue content for the tissue type.
13. The method of claim 1 wherein the anatomical information at the sub-voxel dimensional scale within the voxel volume is adjusted so that the fractional tissue contents of the voxel differ from the fractional tissue contents of the voxel volume by less than the threshold difference by: using a first optimization technique that searches through various position, orientation, and scaling adjustments of the anatomical information at the sub-voxel dimensional scale within the voxel volume to find a set of position, orientation, and scaling adjustments that minimize a squared difference between the computed fractional tissue contents of the corresponding voxel and the fractional tissue contents of the voxel volume; and using a second optimization technique that searches through various warping adjustments of the anatomical information at the sub-voxel dimensional scale within the voxel volume to find a set of warping adjustments that minimize a squared difference between the computed fractional tissue contents of the corresponding voxel and the fractional tissue contents of the voxel volume.
14. The method of claim 1 wherein the anatomical information at the sub-voxel dimensional scale within voxel volumes is adjusted so that the fractional tissue contents of the corresponding voxels differ from the fractional tissue contents of the voxel volumes by less than the threshold difference and wherein the anatomical information at the sub-voxel dimensional scale within voxel volumes is adjusted so that features that span multiple voxel volumes are aligned at the boundaries between adjacent voxel volumes by one or more optimization methods that search through various position, orientation, scaling, and warping adjustments of the anatomical information at the sub-voxel dimensional scale within the voxel volume in order to find a set of position, orientation, scaling, and/or warping adjustments that minimize a squared difference between the fractional tissue contents of the voxel volumes and the computed fractional tissue contents of the corresponding voxels constrained by feature-alignment constraints.
15. The method of claim 14 carried out cluster-by-cluster on clusters of adjacent voxel volumes that need adjustments to the anatomical information at the sub-voxel dimensional scale so that the fractional tissue contents of the voxels differ from the fractional tissue contents of the voxel volumes by less than the threshold difference and so that features that span multiple voxel volumes are aligned at the boundaries between adjacent voxel volumes.
16. The method of claim 14 carried out voxel-volume-set-by-voxel-volume-set for voxel-volume sets, and corresponding voxel sets, containing voxel volumes with identical numbers of tissue types, with voxel-volume sets with lower numbers of tissue types adjusted prior to adjustment of voxel-volume sets with larger numbers of tissue types.
17. The method of claim 1 wherein the anatomical information includes the shapes, positions, and orientations of volumes of each type of tissue present in the organism, or portion of the organism, described by the anatomical information.
18. The method of claim 1 wherein the intensity information associated with each voxel of a medical image includes one or more numerical values, each numerical value selected from a corresponding range of possible numerical values.
19. The method of claim 1 implemented in a computer system.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
February 19, 2019
September 22, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.